MATLAB: To add noise in the image we use J = imnoise(I,​’gaussian’​,M,V) as syntax, why we consider only mean and variance in that syntax to add noise? if variance increase perfomance metrics like psnr decreases and mae & rmse increases why?

Question

To add noise in the image we use J = imnoise(I,'gaussian',M,V) as syntax, why we consider only mean(M) and variance(V) in that syntax to add noise? if variance increase perfomance metrics like psnr decreases and mae & rmse increases why?

Answer 1

You only consider mean and variance because if you've chosen to add Gaussian noise, those two parameters will completely specify the noise. What else do you think you would need? The answer is nothing.

Look at the definitions of psnr, and mae and mse. You'll see that as values get farther away from the perfect reference signal, the errors increase and the signal to noise ratio decreases. It just makes intuitive sense doesn't it? Like if the reference value is 10 but you have noise and so it's 23, then of course the error is high and the PSNR is low. Actually I'm not quite sure where the difficulty in understanding this comes from - it seems so intuitive and obvious.